University of Louisville ThinkIR: The University of Louisville's Institutional Repository

Faculty Scholarship

2-2014

The ultraviolet attenuation law in backlit spiral .

William C. Keel University of Alabama - Tuscaloosa

Anna M. Manning Stennis Space Center

Benne W. Holwerda University of Louisville

Chris J. Lintott Oxford University

Kevin Schawinski ETH Zurich

Follow this and additional works at: https://ir.library.louisville.edu/faculty

Part of the Astrophysics and Astronomy Commons

Original Publication Information Keel, William C., et al. "The Ultraviolet Attenuation Law in Backlit Spiral Galaxies." 2014. The Astronomical Journal 147(2): 14 pp.

This Article is brought to you for free and open access by ThinkIR: The University of Louisville's Institutional Repository. It has been accepted for inclusion in Faculty Scholarship by an authorized administrator of ThinkIR: The University of Louisville's Institutional Repository. For more information, please contact [email protected]. The Astronomical Journal, 147:44 (14pp), 2014 February doi:10.1088/0004-6256/147/2/44 C 2014. The American Astronomical Society. All rights reserved. Printed in the U.S.A.

THE ULTRAVIOLET ATTENUATION LAW IN BACKLIT SPIRAL GALAXIES∗

William C. Keel1,8,9, Anna M. Manning2,8, Benne W. Holwerda3,4, Chris J. Lintott5,6,8, and Kevin Schawinski7 1 Department of Physics and Astronomy, University of Alabama, Box 870324, Tuscaloosa, AL 35487, USA; [email protected], Twitter@NGC 3314 2 Stennis Space Center, MS 39522; [email protected] 3 ESA-ESTEC, Keplerlaan 1, 2201-AZ Noordwijk, The Netherlands; [email protected], Twitter@BenneHolwerda 4 Leiden Observatory, P.O. Box 9513, 2300 RA Leiden, The Netherlands 5 Astrophysics, Oxford University, Denys Wilkinson Building, Keble Road, Oxford OX1 3RH, UK 6 Adler Planetarium, 1300 S. Lakeshore Drive, Chicago, IL 60605, USA; [email protected], Twitter@chrislintott 7 Institute for Astronomy, ETH Zurich,¨ Wolfgang-Pauli-Strasse 27, CH-8093 Zurich, ; [email protected], Twitter@kevinschawinski Received 2013 September 6; accepted 2013 December 28; published 2014 January 14

ABSTRACT The effective extinction law (attenuation behavior) in galaxies in the emitted ultraviolet (UV) regime is well known only for actively -forming objects and combines effects of the grain properties, fine structure in the dust distribution, and relative distributions of and dust. We use Evolution Explorer, XMM Optical Monitor, and (HST) data to explore the UV attenuation in the outer parts of spiral disks which are backlit by other UV-bright galaxies, starting with the candidate list of pairs provided by participants. New optical images help to constrain the geometry and structure of the target galaxies. Our analysis incorporates galaxy symmetry, using non-overlapping regions of each galaxy to derive error estimates on the attenuation measurements. The entire sample has an attenuation law across the optical and UV that is close to the Calzetti et al. form; the UV slope for the overall sample is substantially shallower than found by Wild et al., which is a reasonable match to the more distant galaxies in our sample but not to the weighted combination including NGC 2207. The nearby, bright spiral NGC 2207 alone gives an accuracy almost equal to the rest of our sample, and its outer arms have a very low level of foreground starlight. Thus, this widespread, fairly “gray” law can be produced from the distribution of dust alone, without a necessary contribution from differential escape of stars from dense clouds. Our results indicate that the extrapolation needed to compare attenuation between backlit galaxies at moderate from HST data, and local systems from Sloan Digital Sky Survey and similar data, is mild enough to allow the use of galaxy overlaps to trace the cosmic history of dust in galaxies. For NGC 2207, HST data in the near-UV F336W band show that the covering factor of clouds with small optical attenuation becomes a dominant factor farther into the UV, which opens the possibility that widespread diffuse dust dominates over dust in star-forming regions deep into the UV. Comparison with published radiative-transfer models indicates that the role of dust clumping dominates over differences in grain populations at this coarse spatial resolution. Key words: galaxies: ISM – galaxies: spiral – ultraviolet: galaxies

1. INTRODUCTION UV and optical attenuation measures. Existing models have a wide range in the predicted behavior of both dust mass Our understanding of the effects of dust grains in galaxies has and resulting attenuation with (Calzetti & Heckman increased dramatically with such new capabilities as sensitive 1999), since dust production, destruction, and the shrinking far-infrared (FIR) measures, spatially resolved modeling of the mass fraction in the interstellar medium (ISM) compete in ways spectral-energy distributions (SEDs) of star/grain mixes, and that are not well constrained at large redshifts. Broadly, at- photometry of resolved galaxies giving independent reddening tenuation from SED fits and photometric redshifts has shown maps from the stars themselves (Berry et al. 2012; Dalcanton a peak at z ≈ 1.5, declining at earlier and later epochs et al. 2012). The emerging starlight is modified in ways which (Rowan-Robinson 2003). Observationally, fits to SEDs from depend crucially on the relative distributions of stars and dust, the UV to FIR by Iglesias-Paramo´ et al. (2007) indicate that the and on the small-scale structure in the dust. dust content of low-mass galaxies has increased from z = 0.7 These factors produce proportionally greater uncertainties to the present, while galaxies at high stellar mass show no such in the emitted ultraviolet (UV) range, affecting much of the trend. Deep surveys over wide spectral ranges are now deep and data relevant to galaxy evolution. As extensive Hubble Space wide enough to test models for the evolution of dust. Connec- Telescope (HST) surveys are allowing exploration of the evo- tion of the history of dust mass to observables must fold in at lution of galaxy morphology, they can also help track the least implicit knowledge of the dust distribution on both large evolution of the dust content of galaxies if we can connect and small scales.

∗ A result that has found wide applicability is the effective Based in part on observations made with the NASA Galaxy Evolution extinction law derived by Calzetti et al. (1994), based on Explorer. GALEX is operated for NASA by the California Institute of Technology under NASA contract NAS5-98034. the SEDs of star-forming galaxies. It is relatively flat with 8 Visiting astronomer, Kitt Peak National Observatory, National Optical wavelength as compared to the behavior found from star-by- Astronomy Observatories, which is operated by the Association of Universities star studies, implying that it is strongly affected by the relative for Research in Astronomy, Inc. (AURA) under cooperative agreement with distributions of stars and dust and by unresolved fine structure the National Science Foundation. The WIYN Observatory is a joint facility of the University of Wisconsin–Madison, Indiana University, , in the dust distribution itself. The UV range is particularly and the National Optical Astronomy Observatory. sensitive to these effects, due to the short lifetimes of the stars 9 SARA Observatory. that dominate the UV light from star-forming systems, and to

1 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al. the inevitable bias in favor of more transparent areas within a make it a useful complement to, for example, FIR and submil- finite region of a galaxy (Fischera et al. 2003). Keel & White limeter survey results. (2001) found that that measured reddening behavior in two For most of our studies of dust in backlit galaxies, we backlit spiral galaxies becomes flatter (grayer) when the data have fairly strict symmetry requirements, so we can trace are smoothed over successively larger regions before analysis. radial behavior. For this project, since we are most interested Effects of dust structure must be included in SED models in in the wavelength behavior of attenuation, we can relax this order to retrieve either the intrinsic stellar SED or the effective requirement as long as we can quantify the effects of galaxy extinction. Comparison of galaxy disks seen at various angles asymmetry on our derived attenuation values. Thus, our sample can provide independent information on some of the distribution here includes some galaxies showing mild effects of interactions issues (i.e., Wild et al. 2011). (most notably NGC 2207/IC 2163). This issue highlights the distinction between the observable In this paper, we report an extension of backlighting mea- quantity attenuation, measured on some size scale, and the actual surements into the UV, where most previous attenuation re- extinction attributable to the grain properties, which is typically sults within galaxies are limited either to very nearby systems manifested in what are essentially point-source measurements of (Bianchi et al. 1996; Dalcanton et al. 2012) or to actively star- light from individual stars, where mixture and scattering effects forming galaxies (Calzetti et al. 1994). One motivation for this are negligible (Witt & Gordon 2000). In interpreting observa- study is the promise of using similar techniques in deep HST tions of galaxies, the net loss of direct starlight is often termed images to compare the dust signatures across a significant red- extinction, although attenuation or effective extinction are more shift range, where the typical scaling with wavelength might precise description. Absorption is a property of the grains, while become a dominant factor in comparison with the nearby uni- extinction includes both actual absorption and scattering out of verse. Our results trace the UV attenuation into the outer regions the line of sight; we actually measure attenuation (sometimes of disks, in some cases to areas where the foreground surface described as effective extinction), combining these with scatter- brightness is so low that the properties are not affected by fea- ing into the line of sight, potentially important for large areas tures of the foreground galaxies. Where absolute quantities are within galaxies. These essentially coincide when deriving ex- important, we use the Wilkinson Microwave Anisotropy Probe tinction from individual stars, since scattering from a diffuse “consensus cosmology” (Spergel et al. 2003) values, notably −1 −1 medium can be negligibly small compared to the light of a star. H0 = 72 km s Mpc . For large regions of a galaxy, scattering may be important. Scattering within each galaxy of a pair will be removed by 2. OBSERVATIONS symmetry, so that the asymmetric component of scattering of light from the background galaxy into the line of sight by grains 2.1. Galaxy Sample in the foreground system remains a potential concern. As shown by White et al. (2000), this component drops very rapidly with To measure dust effects using the backlighting approach in the galaxy separation; it also drops with increased clumping of the UV, we need background galaxies that are UV-bright, which in dust (Witt & Gordon 2000). The strong forward scattering of practice means spiral systems (with their attendant lack of exact local grain populations implies that the scattering contribution symmetry). We considered spiral/ pairs from the may drop strongly again at small galaxy separations as the large catalog of overlapping-galaxy pairs generated from Galaxy required scattering angle grows. We note that none of the optical Zoo candidates (Keel et al. 2013). A subset was inspected in images of our targets show the characteristic bluing toward the detail, based on availability of long Galaxy Evolution Explorer edges of dust lanes, which is a sign of scattering (GALEX) exposures (integration >600 s in at least the NUV We have explored the use of backlit galaxies to measure band), large angular size, and suitability for dust analysis based dust attenuation for local systems, with the utility of the tech- on symmetry and the geometry of overlap, as evaluated from nique extended enormously with the production of a catalog higher-resolution optical images. Many of these candidates of nearly 2000 such pairs (Keel et al. 2013) based on exami- proved unsuitable for dust analysis due to low UV surface nation of Sloan Digital Sky Survey (SDSS) images by volun- brightness in the region of interest, lack of symmetry in the teers within the Galaxy Zoo project (Lintott et al. 2008). This UV, or unfortunate location of UV-bright star-forming regions large starting sample allows us to select significant subsamples (identifiable by color as well as brightness). based on galaxy type or geometry. For this study, we concen- trate on systems that are otherwise of limited value in dust 2.2. Data studies—spiral/spiral pairs, where the high UV flux of the 2.2.1. Ultraviolet background spiral makes up for its lack of detailed symme- try compared to E/S0 background systems. Errors in estimating UV imagery came primarily from the GALEX archive the light loss are unavoidably larger than in the optical using (Martin et al. 2005). GALEX carried a 0.5 m telescope, using a E/S0 background galaxies, so we combine data from multiple dichroic beamsplitter to simultaneously observe in both NUV systems and, where possible, average along the spiral pattern to (1925–2730 Å at half-peak) and far-UV (FUV; 1410–1640 Å)  suppress fluctuations due to background structure. A key fea- bands. The resolution is approximately 5.0 FWHM. For some ture of this backlighting approach is that it is weighted by area targets, observed after failure of the FUV detector, only NUV rather than by the luminosity of embedded stars, values which data are available; for others, the total NUV exposures are sig- are clearly more representative for background sources and may nificantly longer than for FUV.These factors, combined with the be most directly representative of the effect in emerging radia- increasingly clumpy structure of galaxies toward shorter wave- tion when averaging over regions of a galaxy several kpc in size. lengths, means our useful sample size shrinks and error bars This approach is explicitly sensitive to a gray component, which grow correspondingly between NUV and FUV bands. In prac- is left poorly constrained by purely spectral techniques, and is tice, we found that only exposures 600 s and longer provided sensitive to arbitrarily cold dust components. These features sufficient signal-to-noise ratio (S/N) for our analysis.

2 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Table 1 Galaxy Pairs Analyzed

System zfg R25 GALEX Exp. (s) Optical Data Models (arcsec) NUV FUV Foreground Background NGC 2207 0.0091 127 8919 13,380 SARA-S Arm tracing Arm tracing NGC 4568 0.0075 137 10,170 1706 HST, WIYN Ellipse Ellipse NGC 5491 0.0197 14 1967 1322 KP2m Symmetry Symmetry UGC 3995 0.0158 27 1535 1535 HST, SDSS Symmetry Ellipse SDSS J143650.57+060821.4 0.0588 19 2485 2485 KP2m Arm tracing Arm tracing SDSS J161453.42+562408.9 ... 20 21,858 6118 WIYN, KP2m Symmetry Symmetry SDSS J163321.48+502420.5 0.0439 21 2698 2698 WIYN Ellipse Mirror symmetry SDSS J211644.67+001022.4 0.0318 16 4513 4513 KP2m Symmetry Symmetry

For the nearby, bright pair NGC 2207/IC 2163, we use An Apogee camera with a 1024 × 1024 Kodak CCD provided data from the XMM-Newton Optical/UV Monitor (OM; Mason a field 10.3onasideat0.61 pixel−1, with ugriz filters. Image et al. 2001). Its 0.3 m primary mirror focuses the image onto quality ranged from 1.9 FWHM at z to 2.1 FWHM at u. a microchannel plate whose output is rapidly read via an For VV 488 = MCG −02-58-11, we use the B and I images intensified CCD, covering a field of 16.2 square at 0.48 pixel−1. from the CTIO 1.5 m telescope reported by White et al. (2000). Relevant UV filters for our program are UVW1 (2450–3200 Å), It proved crucial to have these optical data at substantially UVM2 (2050–2450 Å), and U, with an effective wavelength higher resolution and S/N than the UV images. They show the near 3440 Å. The point-spread function (PSF) for the XMM- context of each system and its level of symmetry (or departures OM is tighter than for GALEX, with FWHM ≈ 3 in the from symmetry) so that we can interpret the UV data with UV filters. Long integrations are available in the NGC 2207 much greater confidence. The optical images were analyzed field: 13,380 s in UWM2 (roughly corresponding to GALEX to yield maps of estimated attenuation at their full resolution, NUV) and 8920 s in UVW1 (somewhat longer in wavelength or discover limiting factors which prevented us from doing than GALEX NUV). The XMM-OM data in B, and to a so. In some cases, despite a favorable geometry of the two smaller extent U, were affected by a reflection artifact from galaxies, the optical data show no attenuation in the apparent an out-of-field source superimposed on the southern part of foreground galaxy (so either this is the background system NGC 2207 (XMM-Newton Community Support Team 2012); or there is very little dust along the backlit lines of sight). the U-band reflection is minor enough that we can analyze Some others are too distorted to apply any of our symmetry this image, but we use the ground-based ug data rather than approaches. Consequently, a comparatively small set of backlit XMM-OM B. Suitable data are not available to extend the galaxies provides our information on the UV attenuation and its analysis of this system to shorter wavelengths; GALEX observed wavelength dependence. These categories of pairs are listed in NGC 2207 in the FUV only during its all-sky survey, and a Tables 1 (for those with UV attenuation measures) and 2 (not long NUV exposure was obtained only after failure of the FUV suitable for such measures; some of these are still suitable for detector. optical analysis).

2.2.2. Optical 2.3. Analysis Methods As set out by, for example, White & Keel (1992), White We have obtained new optical images of many candidate et al. (2000), and Holwerda et al. (2009), the basic technique for galaxy pairs; often the higher resolution and S/N help make the retrieving attenuation from an overlapping galaxy system relies interpretation clear, even when the attenuation measurement for on the expression this project is limited to the resolution of the UV images. When I − F no absorption is detected in the optical images, we infer that the e−τ = , galaxy in question lies in the background of the pair. The new B images came from several sources. where I is the observed intensity at a given point, and F Most images came from the 3.5 m WIYN telescope with and B represent estimates of the intensities of foreground OPTIC fast-guiding camera (Tonry et al. 2002), which provides and background galaxy light without any attenuation (using a10 field sampled with 0.14 pixels. We used this system in three symmetry considerations). We often find it useful to work sessions from 2008 April to 2010 May. Image quality usually in transmission T = e−τ rather than optical depth τ, since ranged from 0.5to0.8 FWHM; for most fields, bright stars the errors from statistics of the data are better behaved and were available allowing us to use the system’s on-chip high- more symmetric than for τ. We neglect effects of scatter- speed guiding. Passbands were B and I; fringing from night-sky ing; the relevant effect comes from the difference in light emission in the I was corrected using median-combined frames from the background galaxy scattered by the foreground be- to generate a reference pattern. For most pairs, exposures were tween the region under analysis and its symmetric point, which 2 × 10 minutes in B and 10 minutes in I. declines very steeply with the line-of-sight separation between The KPNO 2.1 m telescope was used at its f/8 the galaxies (White et al. 2000). Depending on the structure Ritchey–Chretien´ focus, in 2012 April and May. A 2048×2048 of the galaxies and quality of the data available, we use one of TI CCD provided a field of 11 at 0.305 pixel−1, slightly vi- three techniques exploiting different levels of symmetry to esti- gnetted at one edge. Galaxy pairs were generally observed in mate the foreground and background contributions to the light the B (20 minutes) and R (10 minutes) bands. at each point. For the nearby southern pair NGC 2207/IC 2163, we used Ellipse fitting. Here we model one or both galaxies by fitting the SARA-S remotely operated 0.6 m telescope at Cerro Tololo. ellipses at small steps in semimajor axis to the isophotes, then

3 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Table 2 Unsuitable Galaxy Pairs for UV Analysis

System GALEX Exp. (s) Optical Data Problem NUV FUV NGC 4231 11,734 11,734 WIYN Too far apart NGC 4911 1688 1688 WIYN, HST Background too dim in UV NGC 5679 1670 1670 HST, WIYN UV resolution too poor NGC 6365 961 0 KP2m Symmetry failure in UV SDSS J084726.06+533814.9 2922 1696 WIYN Too dim at overlap SDSS J102517.76+170821.0 1722 1722 WIYN Too far apart SDSS J103244.35+543847.5 1680 1680 WIYN Symmetry problem SDSS J105454.32+100250.0 1551 1551 WIYN Background too small/faint in UV SDSS J121326.98+504237.4 2440 1288 WIYN UV surface brightness too low SDSS J121626.29+470131.6 11,774 11,774 WIYN Interacting, distorted SDSS J124415.63+314242.6 3370 1672 WIYN Interacting, distorted SDSS J125725.25+272416.4 31,099 29,932 WIYN Foreground asymmetry SDSS J131222.90+461906.1 3116 1584 WIYN Warped, bad symmetry, faint UV SDSS J131354.20+441048.6 1442 1592 WIYN Background too dim in UV SDSS J131404.57+472145.3 2540 1655 WIYN No absorption detected SDSS J142718.87−014042.4 1680 1680 WIYN Too far apart—arms in wrong place SDSS J154954.44+085140.6 1424 1058 WIYN Background too dim in UV UGC 6212 3213 3213 WIYN Overlap surface brightness too low VV 488 629 0 CTIO 1.5 m Background too dim in UV interpolating to make model two-dimensional images; we use reduces our sensitivity to differences in the normalization of arm the ellipse and bmodel tasks within IRAF/STSDAS,10 which intensity or departures from exact 2θ symmetry in otherwise implement the algorithm from Jedrzejewski (1987). Errors in grand-design spirals. the estimates are generated by propagating the scatter in values Images were aligned using stars in common between optical along the relevant isophotal ellipses for each galaxy, propagating and UV fields (or in a few sparse fields, using galaxy nuclei the standard deviation of these values through calculations of plus scale and orientation parameters known from other fields). transmission and optical depth τ. We rebinned the UV data to the scale of the optical images Point symmetry. Where the galaxies do not have the detailed to avoid loss of information, and smoothed the optical data symmetry needed for the other techniques, we use simple to the resolution of the UV images with the appropriate symmetry by rotating each galaxy image by 180◦, either for Gaussian kernel measured using stars in both image sets. Our a full mapping of the attenuation or measurement averaged over general approach was to use the optical images, with better a resolved region. The error in modeling each galaxy in this way resolution and S/N, to guide our understanding of each pair’s is evaluated by the scatter in apertures of the same size located geometry, and then do identical processing on registered and elsewhere on the relevant isophote for each galaxy PSF-matched optical and UV images to retrieve attenuation Arm tracing. Here we assume that spiral arms have similar values for matching pieces of the galaxies. intensity profiles with radius, using arms in a non-overlapped To avoid potential biases in sample properties, we evaluated region as a guide to the behavior of ridge-line intensity with the suitability of each pair for these measurements before radius and the scatter due to fine-scale structure in the arms as measuring transmission. This was based on geometry, surface evaluated at the relevant spatial resolution. Then the estimate of brightness in the overlapping areas, and absence of UV-bright the non-backlit profile of an arm is taken by scaling the reference knots in a position to distort measures in the overlap area. The arm profile to match non-backlit parts of the arm being analyzed, regions analyzed were selected based on optimal location for allowing intensity scaling. Our technique for doing this starts backlighting (combining background surface brightness with by interactive marking of a number of points on the ridgeline of line of sight penetrating well into the foreground disk) and each arm using a long-wavelength image (I or z). These points evidence of dust from the higher-resolution optical images, to are interpolated in polar coordinates centered on the galaxy avoid contamination by (false) null detections from galaxies that nucleus, sampled typically at 1◦ intervals. Then we can work actually lie in the background. with one-dimensional (resampled) profiles of intensity along each spiral arm as a function of position angle θ. The reference 3. ATTENUATION MEASURES—INDIVIDUAL SYSTEMS arm can be smoothed (typically using a median filter) to reduce the impact of discrete star-forming regions on the profile shape. 3.1. NGC 2207/IC 2163 The error expected in using this as a model of the arm being This pair is a very nearby interacting system. Extinction in the analyzed is taken from the scatter about the mean shape for outer arms of NGC 2207 where they cross in front of IC 2163 multiple independent pieces of the reference arm, of the same was addressed by Elmegreen et al. (2001). Using the spatial angular extent as the dust regions being measured. This approach resolution of HST data, they found diffuse “intercloud” dust along the arms with AV = 0.5 − 1.0, and relatively discrete 10 IRAF is distributed by the National Optical Astronomy Observatory, which clouds with AV = 1–2. The interaction between these galaxies is operated by the Association of Universities for Research in Astronomy has been modeled extensively, incorporating both morphologi- (AURA) under cooperative agreement with the National Science Foundation. STSDAS is a product of the Space Telescope Science Institute, which is cal and kinematic information, with the most prominent distor- operated by AURA for NASA. tion being the ocular form of IC 2163 (Elmegreen et al. 1995).

4 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Figure 1. Montage of NGC 2207/IC 2163 as seen in ugriz and XMM-OM UV bands. Some bright stars have been patched by interpolation in the optical images. The displays use an offset logarithmic intensity mapping, similar to the sinh mapping recommended by Lupton et al. (2004), to retain detail over a wide dynamic range. The area shown spans 254 × 110 arcsec, with north at the top.

This has, so far, left both galaxies symmetric enough to allow starlight contributions of both NGC 2207 and IC 2163 within retrieval of dust attenuation in two arms of NGC 2207, since the this region. We use a 6 × 17 arcsec box with major axis oriented foreground arms are relatively dim where seen against bright 23◦ clockwise from N, centered 3.5 W and 3.1 S of the nucleus parts of IC 2163. of IC 2163. For NGC 2207/IC 2163, we use the XMM-Newton OM data The UV transmissions we measure suggest that the effective in the UV, providing higher spatial resolution than GALEX. covering fraction of clouds with significant optical depth must Figure 1 compares our images of this system from z to UVM1 be much larger than in the optical. This must be so for a realistic (0.95–0.25 μm). Our interpretation is aided by archival HST mix of pixel-by-pixel optical depths, with the result that the WFPC2 images (Elmegreen et al. 2000). The attenuation in cross-section of spiral features is larger than we see in higher- NGC 2207 was measured most accurately using the arm-tracing resolution optical images. We examine this for NGC 2207 using technique to estimate the background intensity and its error, and the archival HST WFPC2 images described by Elmegreen et al. rotational symmetry for the foreground arms in NGC 2207 itself; (2001). We compare distributions of retrieved transmission in for the outermost arm in particular, the foreground intensity is the F336W and F439W filters, approximately U and B bands, so low that its correction makes a negligible contribution to the respectively. error. For the outer arm of NGC 2207 projected against the core Our most detailed results are shown in Figure 2, which of IC 2163 (Figure 3), the foreground-light correction is <20% compared the derived transmission values along the background in this region; we apply a constant surface brightness measured arm of IC 2163 as functions of position angle. We trace the north of the bright disk of IC 2163, and model the background attenuation in three places where the outer arm of NGC 2207 light with a first-order radially symmetric distribution (which is crosses in front of the western (inner) arm of IC 2163 to the especially flat, essentially constant, in F336W). After smoothing northwest (two distinct segments at P.A. −37◦) and one to the over 0.5 due to the limited S/NintheU filter, we derive southwest (P.A. −171◦). The values averaged across these main distributions of relative area versus transmission within the dust foreground-arm crossings are listed with our overall summary lane (in a region comparable to our UV measures) as shown in in Table 3. Errors are based on the scatter in regions with Figure 4. (nominally) no foreground attenuation: θ =−95◦ to −70◦ and θ =−135◦ to −118◦. 3.2. NGC 4567/8 From comparison with other systems, we also evaluated the This spiral pair in the Cluster, despite considerable multiband attenuation using a single box location in the outer overlap in our view, is not strongly interacting. H i mapping, arm of NGC 2207, using simple symmetry to evaluate the as shown by Chung et al. (2009), shows kinematic signs of

5 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Table 3 Summary of Attenuation Results

Foreground Galaxy Type b/a P.A. R/R25 Band Transmission τ (◦) NGC 2207 (box) SAB(rs)bc 0.65 100 0.78 UVM2 0.50 ± 0.66 ... ± +0.85 UVW1 0.63 0.36 0.46−0.45 ± +0.54 OM U 0.64 0.27 0.45−0.36 ± +0.68 u 0.57 0.28 0.56−0.40 ± +0.09 g 0.61 0.05 0.49−0.07 ± +0.09 r 0.67 0.05 0.40−0.07 ± +−.11 i 0.74 0.08 0.30−0.10 ± +0.21 z 0.78 0.15 0.25−0.18 ± +0.47 NGC 2207 region 1 0.76 UVM2 0.16 0.06 1.83−0.32 ± +0.41 UVM1 0.18 0.06 1.71−0.20 ± +0.33 OM U 0.18 0.05 1.71−0.24 ± +0.35 u 0.17 0.05 1.77−0.26 ± +0.12 g 0.27 0.03 1.31−0.11 ± +0.06 r 0.36 0.02 1.02−0.05 ± +0.04 i 0.46 0.02 0.78−0.04 ± +0.04 z 0.56 0.02 0.58−0.04 ± +2.31 NGC 2207 region 2 0.78 UVM2 0.10 0.09 2.30−0.64 ± +1.20 UVM1 0.10 0.07 2.30−0.53 ± +0.62 OM U 0.13 0.06 2.04−0.38 ± +0.43 u 0.20 0.07 1.61−0.30 ± +0.16 g 0.26 0.04 1.35−0.15 ± +0.09 r 0.37 0.03 0.99−0.07 ± +0.05 i 0.43 0.02 0.84−0.04 ± +0.04 z 0.50 0.02 0.69−0.04 ± +0.43 NGC 2207 region 3 0.80 UVM2 0.20 0.07 1.61−0.30 ± +0.25 UVM1 0.27 0.06 1.31−0.20 ± +0.20 OM U 0.27 0.05 1.31−0.17 ± +0.16 u 0.41 0.06 0.89−0.13 ± +0.06 g 0.51 0.03 0.67−0.05 ± +0.03 r 0.63 0.02 0.46−0.03 ± +0.03 i 0.77 0.02 0.26−0.02 ± +0.02 z 0.86 0.02 0.15−0.02 NGC 4568 Sbc 0.39 27 0.73 NUV 0.16 ± 0.15 >1.18 ± +0.15 B 0.51 0.07 0.67−0.12 ± +0.05 I 0.92 0.04 0.08−0.04 ± +0.14 NGC 5491 SBc: 0.81 137 0.68 FUV 0.62 0.09 0.48−0.14 ± +0.13 NUV 0.66 0.08 0.41−0.11 B 1.00 ± 0.03 <0.03 R 1.03 ± 0.03 <0.03 ± +0.12 UGC 3995 Sbc 1.0 0.89 0.95 u 0.70 0.08 0.36−0.11 ± +0.07 g 0.77 0.07 0.26−0.06 ± +0.03 r 0.84 0.03 0.18−0.03 ± +0.04 i 0.85 0.03 0.16−0.04 ± +0.06 z 0.93 0.05 0.07−0.04 SDSS J143650.57+060821.4 Sc 0.58 154 0.75 NUV <1.09 ... B 1.02 ± 0.10 <0.12 +0.05 R 0.890.05 0.12−0.06 ± +0.20 SDSS J161453.42+562408.9 Sbc 0.51 119 0.95 FUV 0.64 0.12 0.45−0.16 ± +029 NUV 0.59 0.15 0.53−0.23 ± +0.30 B 0.76 0.20 0.27−0.22 I 0.84 ± 0.22 <0.47 ± +2.3 SDSS J163321.48+502420.5 SBbc 0.62 75 0.82 NUV 0.21 0.19 1.56−0.58 ± +0.11 B 0.56 0.06 0.58−0.10 ± +0.07 I 0.76 0.05 0.27−0.06 ± +2.6 SDSS 211644.67+001022.4 Sc 0.32 59 0.66 FUV 0.55 0.51 0.60−0.56 ± +0.55 NUV 0.43 0.18 0.84−0.40 ± +0.21 B 0.63 0.12 0.46−0.17 ± +0.10 R 0.72 0.07 0.33−0.12

6 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Figure 2. Slices of derived transmission T along the background arm of IC 2163, as functions of position angle θ. Lines for each filter band are coded in chromatic order, as marked by each curve on the left. Errors are derived from the scatter in apparently absorption-free regions. The black boxes show the location and widths of three regions defined as foreground-arm segments for our analysis.

Figure 3. Region in the outer arm of NGC 2207 analyzed for Figure 4.TheHST WFPC2 F439W image is illustrated over a range 25.6 × 40 around the center of the background system IC 2163; the polygon encloses the pixel area modeled for attenuation to compare covering fractions in F439W and F336W bands. disk warping in the large foreground system NGC 4568; the UV-bright star-forming regions which are common in the inner two disk redshifts coincide quite closely in the overlap region disks of both galaxies. making it difficult to separate the two galaxies’ contributions to The transmission and derived attenuation values are listed in the H i column density in this region. Similarly, the CO velocity Table 3.TheGALEX data give a usable measurement in the NUV field in this region shows a smooth connection, leading Kaneko band; the available FUV exposure is too short for a detection at et al. (2013) to assess the system as being in an early stage of this surface brightness. interaction before first close approach. We measure a dust arm on the north side of NGC 4568, 3.3. UGC 3995 which lies far enough out in the galaxy to minimize effects of stellar spiral structure on the surface brightness (Figure 5). This object, noted in the UGC (Nilson 1973)ashavinga In this area we can use ellipse fits to the background galaxy possible jet, was identified by Keel (1985) as a superimposed NGC 4567 and assess errors from foreground structure from galaxy pair. Both spirals are seen nearly face-on. Their matching the scatter among similarly sized regions on either side of the redshifts indicate potential interaction, but the spiral patterns are dust lane after subtracting the background contribution. This closely symmetric; potential distortions are at larger radii than dust feature, unlike other darker ones, is clear of the luminous, we consider here.

7 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al. signal for measurement in the GALEX bands. We can mea- sure a region of the dust lane closest to the background nu- cleus in all five SDSS filters including u, adding color infor- mation to the previous attenuation values and, for this sys- tem, a bridge wavelength toward the average UV behavior. At these wavelengths, the background light is so dominant that the corrections for foreground light based on the opposite side of the disk are <25%, and the error contributions from these corrections correspondingly smaller. Our transmission maps (Figure 6) show the effect noted by Holwerda & Keel (2013) in which the dust arms lie inward of the bright stellar ridgelines, typically by ≈1.5 or 500 pc.

3.4. SDSS J143650.57+060821.4 We trace intensities along arms in both systems. This pair (hereinafter SDSS 1436 for brevity) is favorable in that the arms Figure 4. Distributions of transmission vs. fractional area in the outer arm of of the background, face-on galaxy are bluer than those of the NGC 2207 from HST images in the U and B bands; the region analyzed for this foreground galaxy, so they are brighter in the UV away from figure is shown in Figure 3. The comparison illustrates the increasing fraction the foreground dust and the error contributed by foreground of the arm area occupied by progressively greater optical depths toward shorter structure is minimized. We have both WIYN OPTIC (BI) and wavelengths. KPNO 2.1 m (BR) optical images, with slightly better surface- brightness sensitivity for the 2.1 m data. The sense of overlap is clear from the light loss against the eastern arm of the western “Snapshot” HST data in the F606W (approximately V) band galaxy where the eastern arm of the other galaxy crosses it; the by Malkan et al. (1998) were analyzed initially by Marziani ≈ more face-on and symmetric galaxy is in the background. et al. (1999), who quote AV 0.18 in the interarm parts of the These arms are nearly tangent to each other in projection; foreground disk, and AV > 1.5 in the arms. Using the favorable we measure transmission in a region where they are closest geometry, with the outer disk of UGC 3995B projected against (Figure 7). Errors are based on the scatter among similarly the bulge of UGC 3995A, Holwerda & Keel (2013) have mapped sized slices of each arm (compared to its symmetric partner) the dust in more detail using the HST image and integral-field in non-overlapped regions. Even if we consider the arm profiles spectra from the CALIFA program (Sanchez´ et al. 2012), noting ignoring the UV-bright knot opposite the overlap region, the a distinct outer edge to strong attenuation and a systematic offset NUV transmission has an error range spanning all physical between the arm locations as defined by stars and dust. values. The bulge component of UGC 3995A is faint in the UV, so we use arm-tracing rather than ellipse fitting for this system. 3.5. SDSS J161453.42+562408.9 To measure attenuation consistently with wavelength, we use this approach for all wavelength bands in this work, rather than The evidence for attenuation of light from the southern adopting the symmetry-based results at high resolution from member of this pair by the northern galaxy is a pronounced HST data by Holwerda & Keel (2013). deficit in the NUV intensity, with a much weaker B deficit As promising as the galaxy types look for UV imaging, the in the same area (Figure 8). We estimate the foreground and background arms are not very blue, so there is not enough background contributions in the overlap region by reflection

Figure 5. Overlap region of the pair NGC 4567/8 shown in the GALEX NUV band and optical BI bands from WIYN OPTIC data. The dust lane analyzed here is indicated. Images are display with an offset logarithmic intensity mapping, where brightness is proportional to log (intensity + constant), approximating the SDSS image rendering. The area shown spans 102 × 36, with north at the top.

8 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Figure 6. Maps of estimated transmission T in SDSS ugriz bands for UGC 3995B. The strong variation in signal-to-noise is driven by the surface-brightness behavior of both foreground and background galaxies. The gray scale ranges from 0 to 2 for the transmission maps, to sample the noise and symmetry artifacts properly. The r image from the SDSS is shown for comparison of the arm patterns. The box on the i transmission map shows the arm region used for attenuation measurements here. The region shown spans 85 × 63.

Figure 7. Galaxy pair SDSS J143650.57+060821.4 shown in the GALEX NUV band and optical BI bands from WIYN OPTIC data. Images are display with an offset logarithmic intensity mapping, approximating the SDSS image rendering. The area shown spans 60 × 73, with north at the top.

Figure 8. Images of SDSS J161453.42+562408.9 with GALEX in the NUV band, and with the WIYN OPTIC system in B and I. Offset logarithmic intensity mapping was used for each to span a wide dynamic range. The box indicates the region in which attenuation was measured. The region shown spans 106 × 115 with north at the top.

9 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Figure 9. Images of SDSS J163321.48+502420.5 with GALEX in the NUV band, and with the WIYN OPTIC system in B and I. Offset logarithmic intensity mapping was used for each to span a wide dynamic range. The circle indicates the region in which attenuation was measured. The region shown spans 106 × 115 with north at the top. symmetry, and evaluate errors from the scatter of multiple Table 4 regions of the same size along the relevant global isophotes Weighted Mean Attenuation Behavior for each galaxy. Sample Band Aλ/AV +0.35 3.6. SDSS J163321.48+502420.5 NGC 2207 UVM2 1.85−0.22 +0.27 UVW1 1.84−0.26 The attenuation in this region appears in front of the northern U 1.76+0.20 part of the edge-on background galaxy. We use symmetry for −0.15 u 1.51+0.17 both galaxies, modeling the NUV image of the foreground −0.13 g 1.16+0.065 galaxy with a linear decline in surface brightness (a close −0.058 r +0.028 fit which allows easy evaluation of scatter about the mean 0.86−0.028 +0.023 behavior). In this system, we measure the attenuation in a i 0.62−0.027 +0.017 single box region (Figure 9), where the error is dominated z 0.48−0.019 +1.12 by foreground structure and evaluated by the scatter along the GALEX FUV 2.71−0.87 overlap isophote elsewhere in the foreground system. +0.75 NUV 2.74−0.40 +0.60 u 1.38−0.53 3.7. SDSS J211644.67+001022.4 +0.12 B 1.24−0.10 We use simple reflection symmetry to model the galaxy light, and evaluate the errors based on the scatter of multiple regions of the same size along the relevant isophote of each galaxy. The system geometry and dust region are shown in Figure 10. Since the data for NGC 2207 are of high quality for each 3.8. NGC 5491 region, we show the values separately (Figure 12). The shaded region encompasses the errors on variance-weighted means in The evidence for attenuation is somewhat ambiguous in this each filter, normalized to V by interpolation between g and r pair, consisting of the drop in surface brightness of the outer bands. arm of the southern spiral where it is projected against the For the systems with GALEX images, we can extend the (apparently) foreground northern companion. Using reflection attenuation data to the FUV band near 1500 Å. The large range symmetry, and evaluating errors from scatter in same-sized in S/N of these measures dictates care in combining data for the regions along the relevant isophotes in each galaxy, our data various galaxies. Some systems have the optical points measured suggest actual light loss, clearly detected in both GALEX UV very poorly, so they would compromise the normalization of the bands but only marginally in the optical bands. The geometry UV values. Accordingly, we normalize the NUV/B ratio via the and region measured are shown in Figure 11. weighted mean of objects with the B values measured at S/N > 1 (in practice, this means S/N > 2.5). Likewise, we normalize the 4. ATTENUATION MEASURES: AVERAGE composite FUV/NUV ratio with a weighted mean of the two BEHAVIOR AND REDDENING LAW systems with non-open error bars (limits for other systems are We combine the results from these galaxy pairs to constrain not informative). the mean attenuation law, connecting light loss and reddening. Figure 13 compares these results to the NGC 2207 means. In doing so, we consider the error ranges as they affect not only While the averages are somewhat higher than for NGC 2207, the weighting of data, but normalization to the optical values. the error bounds (at the standard deviation level) overlap almost The optimum attenuation range in this respect has B values large everywhere, so that at our precision a single attenuation law is enough to be well measured, and the UV values not so large as a plausible fit to the whole data set. The error band is broad to be nearly indeterminate. The combined values are listed in enough for such subtle effects as shifts of effective wavelength Table 4. with reddening to not be of concern.

10 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Figure 10. Images of SDSS J211644.67+001022.4 with GALEX in the NUV band, and with the KPNO 2.1 m imager in B and R. Offset logarithmic intensity mapping was used for each to span a wide dynamic range. The box indicates the region in which attenuation was measured. The region shown spans 60 × 65 with north at the top.

Figure 11. Images of NGC 5491 with GALEX in the NUV band, and with the KPNO 2.1 m in B and R. Offset logarithmic intensity mapping was used for each to span a wide dynamic range. The box indicates the region in which attenuation was measured. The region shown spans 121 × 130 with north at the top.

Figure 12. Wavelength dependence of attenuation in NGC 2207. The normal- Figure 13. Wavelength dependence of attenuation in the GALEX subsample, ization simply interpolates AV between the values for g and r. Points show variance-weighted means, and the shaded region shows the area spanned by the compared to NGC 2207. Combination of data at various wavelengths and S/N standard deviation of these weighted means. levels is described in the text; the u band point comes solely from UGC 3995B. The normalization in the optical simply assumes that AB matches the behavior in NGC 2207. Open circles show variance-weighted means, and the boxed region 5. RADIAL BEHAVIOR IN THE ULTRAVIOLET shows the area spanned by the standard deviation of these weighted means. The error bars are noticeably asymmetric in the UV bands, being symmetric in We can examine the radial behavior of the attenuation, transmitted intensity so that the range extends farther to higher attenuation. mindful of the inherent selection caused by the need to have transmitted light in the UV bands. Therefore, our most accurate greatest. Arm and interarm regions are not clearly distinguished results occur for transmission values not very close to either at the resolution of the GALEX data; only in SDSS 2116 is zero or unity, where the relative effects of symmetry errors are it likely that the UV measure is clearly dominated by arm

11 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Figure 15. Comparison of our mean data, as in Figure 13, to the Calzetti et al. form, using their analytic expression. The assumed value of RV affects both normalization and slope, since our technique can measure gray attenuation as well as reddening. The best-fit value assuming no gray component is AV = 4.0, close to their value of 4.06 from SED fitting.

Optical dust measurements from our data can naturally be performed and interpreted at much higher resolution than their continuation into the UV; we will present such results for a large sample of backlit spiral elsewhere, including both grand-design and flocculent systems.

6. CONCLUSIONS We have used a combination of ground-based optical images Figure 14. Derived optical depth τ as a function of normalized radius R/R25. with GALEX and XMM-OM data in the UV to measure the shape Adjacent bands are plotted together, where the systematic changes are less than the errors in points. Filled circles indicate points identified as being dominated of the effective extinction (attenuation) law in galaxy disks over by dust in individual spiral arms, in NGC 2207 and SDSS 2116. Open triangles the range 1500–9000 Å. Only spiral background galaxies are indicate measurements averaged over larger regions in all other systems. Two bright enough at the UV wavelengths to serve for this technique, points in the B/g plot have been shifted by 0.02 in R/R25 to avoid overlap. so the structure in both galaxies contributes significantly to the error in retrieving the transmission fractions, and subsequently optical depths. The data for the nearby spiral NGC 2207 are dust. Taking the entire sample, the arm regions do show particularly good; since it was also observed with different filters systematically greater attenuation at a given normalized radius than the GALEX system, we consider its attenuation behavior R/R25 than combined arm/interarm values, but our errors separately. are too large to compare attenuation laws in spiral arms The effective extinction law derived by Calzetti et al. (1994) and between them. The high arm values are dominated by from the SEDs of star-forming galaxies has found wide ap- NGC 2207, for which the XMM-OM UV data do resolve the plicability. It pertains to large areas of galaxies, and, like our arms well. results, shows a flatter slope than the intrinsic grain behavior Within these limitations, we summarize in Figure 14 the UV seen in star-by-star investigation of nearby resolved galaxies attenuation as a function of normalized radius, corrected to (Cardelli et al. 1989; Bianchi et al. 1996; Gordon & Clayton equivalent face-on values with a cosine factor (for the assumed 1998). Calzetti et al. (1994) attribute this to a mix of effects: thin-disk geometry, the same as axial ratio b/a). The major the progressive bias in favor of the more transparent parts of (unsurprising) feature of these plots is that data identified as a patchy dust distribution at shorter wavelengths, and the po- arm-dominated show higher attenuation than the other points tential effects of preferential escape of longer-lived stars from averaged across arm and interarm regions. obscuring clouds around regions of recent . Boissier et al. (2005) discuss the radial behavior of attenuation The Calzetti et al. (1994) law is a remarkably good fit to our in the spiral M83 = NGC 5236 from a combination of UV, observations (Figure 15). SED fitting is insensitive to a gray Hα, and infrared tracers. These will likely be weighted toward component of attenuation; if we assume there is no actually star-forming regions, and might be expected to give a higher gray offset, only a single free parameter remains affecting value than our area-weighted data. The outer edge of their both normalization and slope, equivalent to RV = AV /EB−V . relation, at R/R25 = 1.0, has AFUV = 1.1–1.7 mag. For The curve in Figure 15 shows the best-fit value for our data, our sample, the derived face-on values are 0.6–1.9 over the RV = 4.0 ± 0.1, closely comparable to their derived mean = range R/R25 = 0.66–0.95. The UVM2 filter data at a similar RV 4.06. Our results extend the applicability of the Calzetti wavelength for the arms of NGC 2207 also fit broadly with their et al. (1994) form, to even the outer parts of spiral disks where results, A = 1.0–1.5at≈0.8R25. star formation proceeds at very modest levels.

12 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al.

Figure 16. Comparison of our mean data, as in Figure 13, to Wild et al., Figure 17. Comparison of our mean data, as in Figure 13, to the calculations using their analytic expression. Their expression includes stellar mass, axial from Witt & Gordon (2000) for attenuation by both Milky Way and SMC-like ratio, and specific star-formation rate (SSFR) as parameters. We show predicted grains, for various degrees of dust clumping. For each grain type, the upper attenuation relations for the mean axial ratio b/a = 0.61 of our sample, curve is for averaged τV = 0.25, and the lower one for τV = 1. For the entire 10 −9 −1 2 evaluated at extremes of 3 × 10 solar masses and SSFR = 10 yr ,as UV data set, either τV = 1 curve is acceptable in a χ sense, with the NGC 2207 a solid curve, and 3 × 108 solar masses and SSFR = 10−7 yr−1, plotted as a results providing a stronger preference for the same models, implying a strong dashed curve. These are both consistent with the mean for our GALEX sample, role for clumping. For clarity, shaded regions denote ±1σ error bounds for each but not with the UV results for NGC 2207. data subsample.

Particularly in the case of NGC 2207, the foreground-light The GALEX sample by itself is reasonably well fit by the Wild correction and its error are small enough to insure that the et al. curve; however, the errors for these points are substantially slope of the reddening curve we find is not strongly affected larger than for the NGC 2207 data, leaving open the possibility by differential escape of stars from obscuring clouds; fine of a difference. structure in the dust distribution must account for the attenuation This situation clearly represents a sheet geometry, reasonably behavior we find. In nearby, well-resolved spirals analyzed using well approximating the shell geometry used in some radiative- background galaxies, the reddening slope changes with spatial transfer calculations. Calculations have been published includ- resolution in a way consistent with a fractal cloud distribution ing realistic degrees of clumping which we can compare to on scales from tens to hundreds of (Keel & White 2001). our results, for the cases of Milky Way and Small Magellanic This raises at least the possibility that different dust distributions Cloud (SMC) like dust populations (e.g., Witt & Gordon 2000). could occur, giving different reddening behavior, which suggests For this situation, internal scattering effects will be removed caution in applying these results to galaxies at high redshifts. by symmetry, so the more relevant comparison is with the “di- Wild et al. (2011) performed an SED analysis using many rect” component of attenuation excluding scattering. The Witt & pairs of galaxies, matched in specific star-formation rate, metal- Gordon (2000) results, provided in tabular detail, include mul- licity, and axial ratio, to form attenuation curves for subgroups tiple levels of scaling to optical depth averaged over all lines of of galaxies, expressing the results as spliced power-law seg- sight; for a fixed attenuation along one line of sight, this amounts ments varying with each of the matching variables. For their to changing the degree of clumpiness of the dust. We show a results as well, comparison with our mean attenuation data may comparison in Figure 17.Usingasimpleχ 2 figure of merit, the help separate the roles of dust distribution itself from the relative entire UV data set (λ<3000 Å) favors both τV = 1 calcula- 2 2 distributions of dust and stars. Our composite curve is consistent tions (with χ ≈ 1.3) over the τV = 0.25 values (χ ≈ 2); with their UV slopes, steeper than the Calzetti et al. value, over that is, the scenarios with clumpier grains. For the more precise wide ranges in stellar mass and star-formation rate, for the mean NGC 2207 data, the evidence is even stronger, χ 2 ≈ 0.5 versus of the GALEX sample, but not for NGC 2207 (Figure 16). Using χ 2 ≈ 3. The comparison is somewhat degenerate even between their expressions for more inclined disks reduces but does not such extremes as Milky Way and SMC dust, first because the eliminate this discrepancy; our sample has a mean axial ratio NUV and UVM2 bands overlap with the 2175 Å extinction fea- b/a=0.61, with a range of 0.32–1.0 (Table 3). As Wild ture, and second because increased clumping manifests itself in et al. (2011) note, much of the difference they find with axial the calculations as a flatter UV slope in each case, progressively ratio and star-formation rate may trace to distinct populations masking the difference due to grain extinction properties. of grains near star-forming regions and in the diffuse ISM; the Our results suggest the applicability of something very close attenuation statistics with area in NGC 2207 suggest that a pro- to the Calzetti et al. (1994) form for effective extinction of portionally greater fraction of the UV extinction arises in more kpc-scale areas in the outer disks of spirals. This requires a diffuse material. mild extrapolation to the equivalent optical value, indicating that Taken together, our data suggest a flatter UV attenuation further work can use this technique to compare dust content and curve than any of the Wild et al. forms; this could reflect the distribution of galaxies at significant redshifts observed in the nature of backlighting measurements, which are area-weighted emitted UV to the present-day galaxy population as understood independent of the location of stars in the foreground galaxy. in the optical regime.

13 The Astronomical Journal, 147:44 (14pp), 2014 February Keel et al. This project was enabled by many volunteer partici- REFERENCES pants in the Galaxy Zoo project; their contributions to Berry, M., Ivezic,´ Z.,ˇ Sesar, B., et al. 2012, ApJ, 757, 166 the backlit-galaxy program are acknowledged individually at Bianchi, L., Clayton, G. C., Bohlin, R. C., Hutchings, J. B., & Massey, P. http://data.galaxyzoo.org/overlaps.html. This work was sup- 1996, ApJ, 471, 203 ported by the NASA Astrophysics Data Program (ADP) un- Boissier, S., Gil de Paz, A., Madore, B. F., et al. 2005, ApJL, 619, L83 der grant NNX10AD54G. Some of the data presented in this Calzetti, D., & Heckman, T. M. 1999, ApJ, 519, 27 paper were obtained from the Mikulski Archive for Space Calzetti, D., Kinney, A. L., & Storchi-Bergmann, T. 1994, ApJ, 429, 582 Cardelli, J. A., Clayton, G. C., & Mathis, J. S. 1989, ApJ, 345, 245 Telescopes (MAST). STScI is operated by the Association of Chung, A., van Gorkom, J. H., Kenney, J. D. P., Crowl, H., & Vollmer, B. Universities for Research in Astronomy, Inc., under NASA 2009, AJ, 138, 1741 contract NAS5-26555. Support for MAST for non-HST data Dalcanton, J. J., Williams, B. F., Lang, D., et al. 2012, ApJS, 200, 18 is provided by the NASA Office of Space Science via grant Elmegreen, B. G., Kaufman, M., Struck, C., et al. 2000, AJ, 120, 630 Elmegreen, B. G., Sundin, M., Kaufman, M., Brinks, E., & Elmegreen, D. M. NNX09AF08G and by other grants and contracts. This research 1995, ApJ, 453, 139 has made use of NASA’s Astrophysics Data System, and the Elmegreen, D. M., Kaufman, M., Elmegreen, B. G., et al. 2001, AJ, 121, 182 NASA/IPAC Extragalactic Database (NED) which is operated Fischera, J., Dopita, M. A., & Sutherland, R. S. 2003, ApJL, 599, L21 by the Jet Propulsion Laboratory, California Institute of Tech- Gordon, K. D., & Clayton, G. C. 1998, ApJ, 500, 816 nology, under contract with the National Aeronautics and Space Holwerda, B. W., & Keel, W. C. 2013, A&A, 556, A42 Holwerda, B. W., Keel, W. C., Williams, B., Dalcanton, J. J., & de Jong, R. S. Administration. 2009, AJ, 137, 3000 C.J.L. acknowledges funding from The Leverhulme Iglesias-Paramo,´ J., Buat, V., Hernandez-Fern´ andez,´ J., et al. 2007, ApJ, Trust and the STFC Science in Society Program. K.S. 670, 279 was supported by the Henry Skynner Junior Research Jedrzejewski, R. I. 1987, MNRAS, 226, 747 Kaneko, H., Kuno, N., Iono, D., et al. 2013, PASJ, 65, 20 Fellowship at Balliol College, Oxford and by a NASA Ein- Keel, W. C. 1985, AJ, 90, 2207 stein Fellowship at Yale, and gratefully acknowledges sup- Keel, W. C., Manning, A. M., Holwerda, B. W., et al. 2013, PASP, 125, 2 port from Swiss National Science Foundation Grant PP00P2 Keel, W. C., & White, R. E., III. 2001, AJ, 121, 1442 138979/1. Galaxy Zoo was made possible by funding from a Lintott, C. J., Schawinski, K., Slosar, A., et al. 2008, MNRAS, 389, 1179 Jim Gray Research Fund from Microsoft and The Leverhulme Lupton, R., Blanton, M. R., Fekete, G., et al. 2004, PASP, 116, 133 Malkan, M. A., Gorjian, V., & Tam, R. 1998, ApJS, 117, 25 Trust. Martin, D. C., Fanson, J., Schiminovich, D., et al. 2005, ApJL, 619, L1 Funding for the creation and distribution of the SDSS Archive Marziani, P., D’Onofrio, M., Dultzin-Hacyan, D., & Sulentic, J. W. 1999, AJ, has been provided by the Alfred P. Sloan Foundation, the Partic- 117, 2736 ipating Institutions, the National Aeronautics and Space Admin- Mason, K. O., Breeveld, A., Much, R., et al. 2001, A&A, 365, L36 Nilson, P. 1973, Acta Universitatis Upsaliensis. Nova Acta Regiae Societatis istration, the National Science Foundation, the U.S. Department Scientiarum Upsaliensis—Uppsala Astronomiska Observatoriums Annaler of Energy, the Japanese Monbukagakusho, and the Max Planck (Uppsala: Astronomiska Observatorium) Society. The SDSS Web site is http://www.sdss.org/. The SDSS Rowan-Robinson, M. 2003, MNRAS, 345, 819 is managed by the Astrophysical Research Consortium (ARC) Sanchez,´ S. F., Kennicutt, R. C., Gil de Paz, A., et al. 2012, A&A, 538, A8 for the Participating Institutions. The Participating Institutions Spergel, D. N., Verde, L., Peiris, H. V., et al. 2003, ApJS, 148, 175 Tonry, J. L., Luppino, G. A., Kaiser, N., Burke, B. E., & Jacoby, G. H. are The University of Chicago, Fermilab, the Institute for Ad- 2002, Proc. SPIE, 4836, 206 vanced Study, the Japan Participation Group, The Johns Hopkins White, R. E., & Keel, W. C. 1992, Natur, 359, 655 University, Los Alamos National Laboratory, the Max-Planck- White, R. E., III, Keel, W. C., & Conselice, C. J. 2000, ApJ, 542, 761 Institute for Astronomy (MPIA), the Max-Planck-Institute for Wild, V., Charlot, S., Brinchmann, J., et al. 2011, MNRAS, 417, 1760 Witt, A. N., & Gordon, K. D. 2000, ApJ, 528, 799 Astrophysics (MPA), New Mexico State University, Prince- XMM-Newton Community Support Team. 2012, XMM-Newton Users ton University, the United States Naval Observatory, and the Handbook, Issue 2.10, http://xmm.esac.esa.int/external/xmm_user_support/ University of Washington. documentation/uhb/XMM_UHB.html

14